Georeferencing Urban Nighttime Lights Imagery Using Street Network Maps

Abstract

Astronaut photography acquired from the International Space Station presently is the only available option for free global high-resolution nighttime light (NTL) imagery. Unfortunately, these data are not georeferenced, meaning they cannot easily be used for many remote sensing applications such as change detection or fusion. Georeferencing such NTL data manually, for example, by finding tie points, is difficult due to the strongly differing appearance of any potential references. Therefore, realizing an automatic method for georeferencing NTL imagery is preferable. In this article, such an automatic processing chain for the georeferencing of NTL imagery is presented. The novel approach works by simulating reference NTL images from vector-based street network maps and finding tie points between these references and the NTL imagery. To test this approach, here, publicly available open street maps are used. The tie points identified in the reference and NTL imagery are then used for rectification and thereby for georeferencing. The presented processing chain is tested using nine different astronaut photographs of urban areas, illustrating the strengths and weaknesses of the algorithm. To evaluate the geometric accuracy, the photography is finally matched manually against an independent reference. The results of this evaluation depict that all nine astronaut photographs are georeferenced with accuracies between 2.03 px and 6.70 px. This analysis demonstrates that an automatic georeferencing of high-resolution urban NTL imagery is feasible even with limited attitude and orbit determination (AOD). Furthermore, especially for future spaceborne NTL missions with precise AOD, the algorithm’s performance will increase and could also be used for quality-control purposes.

Keywords

• nighttime remote sensing
• NTL
• street network map
• georeferencing
• image matching